The rate of obsolescence of knowledge in software engineering | Hacker News Reader

The rate of obsolescence of knowledge in software engineering

daltyboy11.github.io

34 pointsdgs_sgd17 comments

"Software engineers with a traditional computer science background learn things that never expire with age"

Well they do to an extent, because you use them so rarely that you forget.

Software Engineering is a craft and much of it is timeless. Unfortunately the hiring process revolves around trendy frameworks / and tools.

> Unfortunately the hiring process revolves around trendy frameworks / and tools.

Yeah, this. Sometimes I envy the leetocde way of hiring in the US, since at least that test's your brain and craft and not what frameworks and technologies you used like they do in Europe:

Recruiter: "My client is looking for an experienced dev with Node and AWS experience. What's that, you have 15 years of industry experience but it's all on-prem Java/PHP? Too bad. NEXT!"

"I'm looking to fill a DevOps position for a client and your profile caught my eye. How many years of experience do you have with Linux and Ansible?"

It's sad that in order to get to a technical interview you have to pass the BS recruiter filter consisting of years in X tool/framework instead of skill and domain knowledge.

> like they do in Europe

That's a pretty broad statement. I've had many styles of interviews in Europe; from whiteboard tests, leetcode tests, domain specific tasks, to chatty lunches with no practical test.

The interview yes, it's like how you said, but usually before you get to an interview you have to pass a BS filter from a less technically savvy recruiter who skims through your resume looking for the years of experience in X tool/frameowrk the hiring manager told them he uses.

I just got done hiring a team, and I won't touch leetcode as an indicator for candidates. I select for attention to detail, cognizance of knowing what they don't know, predilection to running things down, and an understanding of how things can go wrong without escalating. Also I look for an appreciation for distinguishing what is enough to get by vs the impractical but perfect. Leetcode teaches or creates as a signal none of that.

In fact I despise Leetcode as both candidate and consumer. It caters to memorization instead of understanding, and is heavily biased toward those that will unquestionably just vomit out what they're told instead of really getting down to understanding what is getting asked for.

Toy problems are cute, but disrespectful of just about every worthy candidates time, and a waste of my time when what I'm looking for is a human capable of learning, contributing, and communicating; not just being a replacement cog in a machine.

I guess it depends on what kind of developer you are. You might do easy stuff in trendy frameworks, in which case you're closer to the physicist - but only in the sense that your knowledge is rendered obsolete every couple of years.

If you operate deeper, you're much closer to the English professor. The fundamentals don't really go thru revolutionary changes.

I agree. My goal in my own career would be to operate on that deep level and be more like the English professor, but so far it’s been more like the physicist

That's true, but occasionally revolutions happen.

"artificial intelligence" and "machine learning" were pretty static from the early '1990s to the late '2000s - and then, all of a sudden there was an explosion which many "English professors" missed.

The ability to debug at a system level never goes away. I’m finding that the lessons I learned 20 years ago remain relevant, only the tools to gather and analyze the data have changed.

> What would the half-life of the value of a software engineer’s knowledge be? I suspect it’s somewhere between the physicist’s and the English professor’s because a software engineer’s knowledge is a combination of eternal computer science/engineering principles and ephemeral technologies that drift in and out of popularity over time.

Isn’t most of a physicist’s knowledge eternal, relative to a software engineer’s? After all, software engineering is based on systems designed by man while physics is modeling the fundamental universe.

In software engineering a better term would be end-of-life. Vogue frameworks and paradigms are more about building communities. Exciting communities help people who are getting started to learn as well as staffing and conferences. What is and isn't popular has very little to do with engineering. Take into consideration that 1/3rd of all programmers, probably only started doing it in the last couple years. https://archive.is/cSzqy

You're getting downmodded somewhat unfairly imo. People here underestimate heavily the momentum of fads in our industry. So many technologies that the industry rallied behind fell by the wayside purely because something newer displaced it by having the energy and momentum of youth behind it. After close to three decades in the industry it becomes very obvious when and how the wind of change is blowing and what's being swept by the wayside.

Author here! I came across the original study referring to the "physicist" and the "English Professor" while reading Economic Facts and Fallacies by Thomas Sowell. It's an interesting read that addresses almost all of the hot political topics of today.

I’ve heard a lot about Sowell, particularly in terms of his ability to frame today’s current events in a prescient way, but haven’t read anything from him. Do you have a particular work from him that you’d recommend as an intro/starter?

The first book of his I read was Intellectuals and Race. Economic Facts and Fallacies is similar in terms of how the arguments are framed (though it touches on different subjects). I'm not sure if this is true of all his books but I would recommend looking through his complete works (there's a lot) and picking one on a topic you're particularly interested in :).

See above

https://youtu.be/V6ZPg6kOBkc

Sowell discussing common falacies about race and economics

Reply on news.ycombinator.com